- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
06-02-2021 05:37 PM
Where exactly is the MLFlow Tracking Server that is managed by Databricks located? Is it provisioned on the same instances as the Databricks cluster (ie. is it part of the EC2 cluster, or is it some standalone service )?
Accepted Solutions
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
06-17-2021 04:30 PM
It's run in the Databricks 'control plane', not within a particular cluster, and not within the customer's account. Databricks manages it and the metadata it stores, but artifact data is still stored in DBFS and thus in the customer's storage account.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
06-17-2021 04:30 PM
It's run in the Databricks 'control plane', not within a particular cluster, and not within the customer's account. Databricks manages it and the metadata it stores, but artifact data is still stored in DBFS and thus in the customer's storage account.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
06-22-2021 04:47 PM
The previous answer is applicable for managed MLflow as part of Databricks Machine Learning.
For Open Source MLflow please see the 4 different scenarios described in the Open Source MLflow website https://mlflow.org/docs/latest/tracking.html#how-runs-and-artifacts-are-recorded.